1 HIV Viral Load Working Group

Spain Jordi Casabona for the PISCIS Study Group Center d'Estudis Epidemiologie sobre ITS/VIH/SIDA de Catalunya. Ronald Bosch for the ACTG trial group. This study was supported in part by the AIDS Clinical Trials Group funded by the National Institute of Allergy and Infectious Diseases (AI 68636, AI 38858, AI 68634, AI 38855). In addition, a growing body of in vitro evidence suggests that co-infection with malaria [17, 18] and tuberculosis [19-21] can lead to significantly higher HIV replication rates.

This increase in genital HIV-1 viral load in plasma RNA observed with HSV-2 co-infection places this STD in a unique position compared to other non-STD co-infections such as malaria and tuberculosis. We studied the effect of geographic region on viral load using data from 25 studies of 44 subcohorts of HIV-infected, antiretroviral-naïve patients in 7 geographic regions. CD4 categories may contain VL measurements for different individuals or for the same individual at different time points.

We considered the effects of the following factors on log10 viral load: region (North America, Europe, Asia, South America, East Africa, West Africa, and South Africa) or country (South Africa); CD4 category; pregnancy; sex; HIV subtype (the predominant subtype in the local region); and test (the main viral load test used in the study). Future prospective studies would require the use of the same viral load test to compare VL between regions. The model weights each observation based on the sample size in that category in the study (i.e., CD4 category by sex category by pregnancy status).

The regional differences in mean VL compared to North America are shown in Figure S3.1.

Sensitivity analyses on the statistical modelling

Discrepancy of the data from South Africa

The removal of these three sub-cohorts left two South African sub-cohorts for which the PCR test was used.

Figure S3.2: Differences in log 10 VL between the bDNA V 3.0 and Roche Amplicor HIV-1 MONITOR (HIM) 1.5 PCR assay among subtype C HIV infections in South Africa [62]

Model equations and description

The rates S iHIV( ) are the forces of HIV infection (risk rates of infection .) experienced by susceptible populations (S i( )and BS i(. Note that we use the term effective rate of partner switching, in compared to the rate of partner change, since this parameter does not simply reflect the actual rate at which individuals change their partners, but also represents other behavioral mechanisms that effectively increase this amount, such as alignment and topology of sexual networks [66-68] and variability in risk Behavior [69] The mixture between the four risk groups is dictated by the sexual mixture matrix G( , )i j which provides the probability that an individual in risk group i would choose a partner in risk group j[70].

At one extreme, e0, mixing is fully proportional, while at the other extreme e1, mixing is fully assortative, as individuals choose partners only from their risk group. The parameter qMorb describes the fractional decrease in coital frequency due to the presence of the elevated VL biological cofactor (to account for co-infection morbidity, because the increase in VL is potentially due to co-infections). We assume a constant birth rate in the model and we do not stratify the population by gender.

The best data on the risk of HIV transmission by sexual act indicate that, in the absence of male circumcision, there are no differences between male-to-female and female-to-male HIV transmission rates by sexual act [72, 73] . The causes of the disparity in HIV prevalence among men and women seem to be mainly due to behavioral patterns such as the mixing of age cohorts (young women with older men) [74].

Biological and behavioral input to the model

The duration of the acute, latent and late stages is believed to be 49 days (acute), 9 years (latent), and 2 years (late). 73], Pinkerton's analysis of the Rakai data for acute infection [77], the measured time from seroconversion to death in the Masaka cohort in Uganda [78], recent findings in the Mombasa cohort in Kenya [79], and recently . The population is divided into four sexual risk classes: low, low to moderate, moderate to high and high risk groups.

Populations with more than one non-spouse, one non-spouse, and no non-spouse partnership in the previous year characterize intermediate to high, low to. The duration of the sexual lifespan is set at 35 years to fit the age groups 15–49 years typically used to define the sexually active population by the WHO as well as many HIV studies [83, 84]. We used the general population survey from the Four City Study [83] to match HIV prevalence levels in Kisumu, Kenya, the representative setting we used to assess the impact of the VL effect; and prenatal clinic surveillance data [ 85 , 86 ] to fit the time series trends in HIV prevalence.

Logarithmic increase in HIV viral load level in sub-Saharan Africa due to HIV plasma viral load infectivity-enhancing effect. Low to medium risk 1.3 partners/year Ibid Medium to high risk 5.5 partners/year Ibid High risk 43.0 partners/year Ibid. Low risk with low to medium risk 36 months Ibid Low risk with medium to high risk 24 months ibid Low risk with high risk 12 months Ibid Low to medium risk with low to.

Intermediate to high risk with high risk 1 months Ibid High risk with high risk 1 week [89]. HIV transmission probability per coital act in the presence of an increased VL effect of size 0.58 log10. Percentage increase in HIV transmission probability per coital act in the presence of an increased VL effect of size 0.58 log10.

Probability of HIV transmission by sex act in the presence of an increased VL effect of 0.46 log10. Percentage increase in the chance of HIV transmission per coitus in the presence of an increased VL effect of 0.46 log10. Incidence of HIV due to HPVLIE effect at that time Total HIV incidence at that time.

Table S4.2: Effect of heightened VL on HIV transmission probability per coital act

Further results on modeling the epidemiological impact of higher HIV-1 plasma RNA viral loads in sub-Saharan Africa

Impact of the VL effect per risk group

Impact of the VL effect assuming 20% faster disease progression with heightened VL

Impact of the VL effect at different levels of the heightened VL Figure S4.3 shows the impact of the VL effect on the population attributable fraction at

Impact of the VL effect at different levels of sexual activity reduction with heightened VL